Pareto analysis for gene filtering in microarray experiments

We introduce a method for detecting strongly monotone evolutionary trends of gene expression from a temporal sequence of microarray data. In this method we perform gene filtering via multi-objective optimization to reveal genes which have the properties of: strong monotonic increase, high end-to-end slope and low slope deviation. Both a global Pareto optimization and a pair-wise local Pareto optimization are investigated. This gene filtering method is illustrated on mouse retinal genes acquired at different points over the lifetimes of a population of mice.